The corpus has 83,911 live India tech postings, and the typical candidate applies to a vanishingly small slice of it — on the order of 0.1% (roughly 80–90 jobs) — usually chosen by keyword and panic, not fit. Spraying applications feels productive and produces almost nothing, because you are competing on the most crowded, least-matched listings. The cluster-aware approach flips it: instead of more applications, you make fewer, sharper ones aimed at the skill cluster you actually belong to.
Why spraying fails
When you apply to 100 loosely-matched jobs, three things happen: your materials are generic (no time to tailor), you land in the highest-volume listings (worst odds), and you never build a signal with any one cluster of employers. The result is a low reply rate that you then try to fix with more volume — the exact wrong move.
| Approach | Apps / week | Tailoring | Typical outcome |
|---|---|---|---|
| Spray | ~100 | None | Low reply rate, burnout |
| Cluster-aware | 10–15 | High | Higher reply rate per app |
What a skill cluster is
A cluster is a group of skills that keep appearing together in postings — e.g. "Kubernetes + Terraform + observability" is a platform/infra cluster, while "LLM integration + vector DBs + Python" is an AI-platform cluster. Employers hire for clusters, not for isolated keywords. If your skills sit inside one cluster, the postings in that cluster are your real market — and they are a far smaller, more winnable set than "all tech jobs."
The cluster-aware loop
- Locate your cluster. Map your top 4–5 skills onto the live cluster map and find the cluster they anchor.
- Filter to it. Restrict your search to postings inside that cluster, in your target cities and seniority tier.
- Tailor deeply. Because the set is small, you can write a real, specific application for each one.
- Track outcomes by cluster. If replies are low, you may be one cluster off — adjust, don't just add volume.
The maths of fewer, better
The point is not effort for its own sake. Ten tailored applications to your own cluster routinely beat a hundred generic ones, because reply rate per application rises far more than enough to offset the lower count — and you finish the week with energy left over. (The reply-rate figures vary by person and market; the structural advantage of matching does not.)
| Week of… | Apps | Tailored | Energy left |
|---|---|---|---|
| Spraying | 100 | 0 | None |
| Cluster-aware | 12 | 12 | Plenty |
Start here
Find your cluster on the skill-cluster map, check what your matched roles actually pay on the salary bands explorer (and learn to read those numbers via how to read a salary band), and target the durable end of the market using the most defensible tech jobs in India. Questions: [email protected].
Methodology & data sources
Figures in this piece are computed from Tevos Labs' job-market corpus — a continuously-crawled, deduplicated index of India tech postings. Aggregates were read on 2026-05-28 from the following tables:
rich_job_enrich— 83,911 live postings used for the corpus size figureskill_clusters— co-occurrence clustering that defines each clusterapplication_events— anonymised apply volume used for the ~0.1% coverage estimate
Salary distributions are computed only where a band has a sample size of n ≥ 12 postings; smaller cells are suppressed. Percentiles (p25/p50/p75/p90) are empirical order statistics over observed advertised ranges, not modelled estimates. Numbers marked "illustrative" are worked examples for explanation, not corpus reads. Questions on method: [email protected].